In this paper we analyse the extended Amdahl's law metioned in the papers [16-19] and demonstrate that data preparation is another promising approach to improve performance of heterogeneous system.
Section II revisits the extended Amdahl's law of asymmetric multicore system by considering the overhead of data preparation [13].
Revisiting the extended Amdahl's law by considering the overhead of data preparation
In this section, we describe the analytic modeling techniques based on Amdahl's law for multicore architectures.
Hill and Marty [10] did one of the pioneer works by extending Amdahl's law to multicore architectures by constructing a cost model for the number and performance of cores that the chip can support.
It seems that someone is again trying to help me to "understand Sandia's results," resolving the apparent contradiction between Amdahl's law and our experimental observations.
The original Communications note did not use primes to clarify that s and p change meaning in going from Amdahl's law to Sandia's, although it is explained in the text of the note.
In the symmetric load balancing, the performance is limited by Amdahl's law.
8, the proposed load balancing approach can provide better performance than the performance obtained by Amdahl's law with typical symmetric load balancing.
The main thrust of the technical note was that Amdahl's law lacks predictive value.
Amdahl's law only tells us what will happen for a fixed problem size, and we want to know what would happen in the practical situation where the problem size grows with available computing power.
This reasoning gives an alternative to
Amdahl's law suggested by E.
